Three-stage convolutional neural network-based object grabbing detection method

Abstract

The invention discloses a three-stage convolutional neural network-based object grabbing detection method. The method comprises the following steps of: 1, obtaining a data set; 2, constructing networkstructures of first-stage, second-stage and third-stage convolutional neural networks, and training the convolutional neural networks; 3, obtaining pre-selected grabbing boxes of a target object andjudging values of the pre-selected grabbing boxes by utilizing the trained three-stage serial convolutional neural network; 4, obtaining an optimal grabbing box through the judging values; and 5, determining a position and a posture of the target object. The method is capable of improving the correctness of the grabbing boxes and realizing high-correctness grabbing for unknown objects.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to an object grasping detection method based on a three-level convolutional neural network. Background technique [0002] As the basic function of robots, object grasping operation has always been an important research direction in the field of robotics. In order to improve the success rate and accuracy of object grasping, many researchers take the grasping point of the object as the research object, and improve the success rate and accuracy of grasping by selecting the best grasping point of the object. Before the introduction of deep learning, in most cases, the grasping point of the object was determined by manually designing features or through the three-dimensional model of the object. The object of the 3D model is known, but the grabbing point of the unknown object cannot be obtained. After the introduction of deep learning, although the convolutional neu...

AI Technical Summary

Problems solved by technology

Before the introduction of deep learning, in most cases, the grasping point of the object was determined by manually designing features or through the three-dimensional model of the object. The object of the 3D model is known, but the grabbing point of the unknown object cannot be obtained

After the introduction of deep learning, although the convolutional neural network can be used to detect unknown objects, the accuracy of the grasping point is relatively low and needs to be further improved. Therefore, it is necessary to further improve the method of obtaining the best grasping point so that the object Grasping is not restricted by unknown objects, and has a high success rate and accuracy of grasping

Images